Movements of the visual scene elicit ocular following at short latency. Our previous work had suggested that the underlying tracking system may help stabilize gaze in a world with three-dimensional structure by responding preferentially to images in the plane of fixation. Accordingly, we have now tested the dependence of early ocular following on binocular horizontal disparity in two humans and two monkeys. Subjects faced a tangent screen onto which two identical random dot patterns subtending 80 degrees x 80 degrees were back-projected. Orthogonal polarizing filters in the two projection paths ensured that each of the two eyes saw only one of the patterns, movements of which were achieved with mirror galvanometers. Stimuli were step-ramps applied in the wake of 10 degrees centering saccades: steps were disconjugate (disparities ranging from 0 degrees to 12.8 degrees) and served to position the scene with respect to the plane of fixation, whereas ramps were conjugate (40 degrees /s, leftward and rightward, with durations of 100 milliseconds (msec) for monkeys and 200 msec for humans) and served to elicit ocular following. Shutters in the two projection paths allowed a brief blanking (lesser 15 msec) of the images during the onset of the step-ramp to eliminate any differential smear due to the step; such blanking had little effect on early ocular following. Ocular following of all four subjects showed strong modulation with disparity, with effects on initial eye acceleration and latency, mostly the former in monkeys and the latter in humans. Initial eye accelerations were greatest (and latencies shortest) when the scene was imaged in the immediate vicinity of the plane of fixation. Eye accelerations decreased and latencies increased when the scene was imaged outside the plane of fixation, these effects reaching a maximum with a disparity of 3.2 degree (crossed or uncrossed). In some cases, there was slight recovery with further increases in disparity (Mexican hat vs. bell profile), possibly reflecting active inhibition outside the plane of fixation or default responses to uncorrelated images. In sum, the data indicate the motion detectors mediating early ocular following in both monkeys and humans are selectively sensitive to images in the immediate vicinity of the plane of fixation.